Mapping RNAseq data to genome from related organism

[hlicht]

Hi,

I have RNAseq data sequenced with paired-end 100bp illumina from a fungus with no reference genome. However, we have sequenced the genome of a closely related species and is running gene predictions for this genome at the moment.

I was wondering what the best strategy for assembling a transcriptome for the species without a reference genome. I thought of using STAMPY (or maybe TOPHAT with appropriate mismatch settings) to map the sequence reads to the genome of the related species, as an alternative to a de-novo assembly with foreaxmple TRINITY. Does anyone have experience with this? Is STAMPY a good choice?

It of course boils down to how closely related the organisms are if it will work. Judging from coding regions of single genes between the two organisms there is only ca. 5% basepair differences per gene, but more basepair differences and indels in the non-coding regions.

Many thanks in advance for any thoughts and recommendations.

[Brian Bushnell]

I won't comment on the best strategy for assembling a transcriptome.

But if you want to use mapping, BBMap works for RNA-seq data and is much more sensitive than Tophat (particularly for indels), and thus better for cross-species mapping. It will work fine with the default settings:

bbmap.sh in1=read1.fq in2=read2.fq out=mapped.sam ref=fungus.fa nodisk

[westerman]

The 'Trinity' program is often used for assembling a transcriptome. There is a 'genome guide' option for when you have a closely related assembly.

[hlicht]

Many thanks for your replies.

BBMap looks very interesting, I'll definitely see how it performs on my data.

Also thanks for pointing out the genome-guided trinity option, I wasn't aware of that. It looks like exactly what I'm looking for.

Thanks again for your replies.

[whataBamBam]

Hi hlicht

Are you still around? How did you get on with this?

I'm in the same situation - Rna-Seq of a fungus with no reference, but we have the genome of a close relative which I'm using for mapping reads.

I'll try out BBMap as well, but would be interested to compare notes.

Tophat with default paramters performs badly (2% of input mapped)
Increasing mismatch and edit distance parameters to 7 leads to 20% of input reads being mapped.

I'll try relaxing tophat's parameters a little more and compare to BBMap.

Anyone else have experience of this? Noone else using closely related genomes to do mapping based Rna-Seq? Everyone going de-novo instead?